A Study of Hand Gesture Recognition Technique Engineering College, Mouli (Barwala) ... resolution images has always been the focus in the dispensation of the digital images.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056

Volume: 03 Issue: 06 | June-2016 www.irjet.net p-ISSN: 2395-0072

© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 1443

A Study of Hand Gesture Recognition Technique

Er. Garima Baweja1

Electronics & Communication Department, Panchkula Engineering College,Mouli (Barwala)

Er.Navjot Kaur2

Assistant Professor , Electronics & Communication Department, Panchkula Engineering College, Mouli (Barwala)

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract : Hand gesture is the method of the hand

movement or the way in which we can identify a hand of an

individual and recognition of an individual. Hand Gesture

Recognition is the biometric process by which an individual

can be identify by the motion of his hand. Hand gesture

recognition has a number of applications in feature extraction,

machine vision, virtual reality, machine manage in industry

etc. In this paper we present the review of hand gesture

recognition method and different approaches SVM, Speeded up

Robust Feature (SURF).

1.INTRODUCTION Recognition of an individual is an

important task to identify people. The detection through

biometric is a better way because it relate with individual

not with in order passing from one place to another.

Biometrics was about perceiving those naturally. Basically

we verify a set of numbers that are appealing to a specific

individual. The definition of Hand Gesture is defined as

“Movement of our hand”. Hand gesture as a biometric

recognition technique used to study in the domains of work

station vision. It has developed advantages inside the

workstation vision group and various stride measurements

have been produced. Hand gesture recognition is an

increasing biometric innovation in which individuals are

absolutely recognized by the movement of their hand. It has

been pulled in advancement as a technique for ID on the

grounds that it is not obtrusive and does not oblige the

subject's participation. Hand gesture recognition could be

utilized from a separation that making it appropriate to

recognizing the culprits at a wrongdoing scene. The hand

gesture of an individual could be caught at a separation of

dissimilar to different biometrics. For example: - fingerprint

recognition.

Hand gesture recognition works from the sensitivity that a

singular's strolling style is one of a kind and could be utilized

for human distinctive proof. In bank situation, just few

accepted individuals are allowed to go into lockers room,

here tread examination system is utilized, hand gesture

movement successions of those approved individuals are put

away in bank's database, thusly at whatever point an

unapproved individual tries to go into room, his movement

of hand won't match with put away groupings and alert

framework will be enacted for any activity.

1.1 HAND GESTURE RECOGNITION SYSTEM

This method includes various methods for recognition:

Feature Extractions: This is an essential step in hand

gesture recognition. The feature must be robust to in use

conditions and should yield good discriminability across

individuals. Each hand motion sequence is divided into

cycles. Hand gesture cycle is define as person starts from

rest, left hand forward, rest, right hand forward,

arrangements rest. The stance during hand gesture cycle,

Hand gesture cycle is determined by conniving sum of the

foreground pixels. At rest position this value is less. By

computing range of frames between two rest positions, hand

gesture cycle (period) is estimated.




1.2 Matching and Recognition:

Matching and Recognition is the final step of hand gesture-

based person identification. Here, input test video sequence

is compare with the trained chain in the database. In general,

minimum distance classifier may be used for hand gesture

recognition. In the training, after parallel processing of two

training processes, spatial and temporal templates are

extracted. Test sequences are pre-processed by template

extraction and projection.

1.3 Model based approach: Model-based

methodologies utilize models whose parameters are

controlled by handling of stride groupings (paired shapes).

These systems are scale, view invariant and oblige great

quality feature arrangements. In this methodology human

outline is partitioned into neighbourhood areas relating to

distinctive human body parts, and ovals are fitted to every

area to speak to the human structure.

HANAVAN MODEL: The statistical human body model

planned by Hanavan’s. This was initially try by Miller &

Morrison. The stalk was divided into three segment at the

omphalion (navel) and xyphion level upper (elliptical

Column), middle (elliptical solid) and lower (elliptical

column). The hand was distinct as an ellipsoid of revolution.

The foot was defined as an elliptical solid with base

(proximal end) being circular. The thigh was defined as an

elliptical solid with top (distal end) being circular. A total of

41 anthropometric parameters need to be measured in this

model.

Speeded Up Robust Features (SURF features) is a

vigorous local feature identifier. It was representing by

Herbert Bay .that could be utilize within machine vision

activities like item distinguishment or 3D imitation. It is

partly encouraged by the filter descriptor. The standard

demonstration of SURF is a duo times snappier than SIFT

and ensured by its inventors to be more able against

conflicting picture transformation than SIFT. The most

precious property of a concentration point detector is its

repeatability. The repeatability expresses the dependability

of a detector.

Support vector machine (SVM) The Support Vector

Machine is a state-of-the-art classification method .The SVM

classifier is usually utilize as a part of bioinformatics (and

different orders) because of its intensely exact, ready to

figure and procedure the high-dimensional in sequence, for

example, gene interpretation, and edibility in display

assorted well springs of data . SVM is fit in with the common

class of piece systems. A piece system is a computation that

relies on upon the information just through spot items. At

the point when this is the situation, the dab item could be

supplanted by a bit capacity which registers a spot item in

some possibly high dimensional peculiarity space. This has

two points of concern: First, the ability to create non-direct

choice limit utilize routines intended for straight classifiers.

Second, the utilization of bit capacity permits the client to

apply a classifier to Data that have no evident settled

dimensional vector space demonstration. The double SVM

issue give us a chance to find the supreme idea of vector.

Identically, the compare qualities are non-zero.

Consequently, the help vectors are the "vital" prepare

focuses, and the purpose of preparing is to uncover them.

2. LITERATURE REVIEW

This block described the research work that has been done

in recent years. Image compression is the ultimate

favourable field of research in which assemble the interest of

all analysts. A literature review goes beyond the inquiry of

report or knowledge and it relates the recognition and

connection of relationships among the literature and

research field.

D.K. Vishwakarma,Rajiv Kapoor and Rockey Maheshwari

“et.al” [1] —In this paper, a simple and effective move




toward for the recognition of hand gestures from very low

resolution images was projected. Improvement of the low

resolution images has always been the focus in the

dispensation of the digital images. Images with declaration

as low as [50×50 pixels] are also taken for recognition. The

gestures under thought here were the number of fingers

(one, two, three, four or five) increased by the person. The

less resolution gesture picture capture from mobile phone,

web camera, or low cost cameras was processed

methodically to amount produced the number of fingers

raised. Simple logic of the geometry of the hand has been

used for the identification of hand gesture from the input

low declaration images. The projected method extracts the

hand gesture directly from the low resolution image without

the need of renovation to a high resolution image or use of

any classifier. The proposed method is based on the creation

of a mask for the image which was vital in the recognition of

the hand gesture recognition.

Dhanashree Pannasa “et.al” [2] Almost all purchaser electric

apparatus equipment today utilizes isolated controls for user

interfaces. Although, the type of individual types and focused

directions that each isolated order distinctiveness

furthermore raise many Problems: the adversity in locating

the needed inaccessible command, the disorder with the

button design, the substitution topic and so on. The buyer

electronics domination design utilizes hand signs was a new

inventive client interface that resolves the problems of using

many inaccessible controls for household machines. We

advise such a method for automatically identify a restricted

set of signs from hand resemblance for electronics

equipment command purpose by means of straddling

consecutive facts and figures outcome from PC to wireless

device manager circuits. Hand gesture recognition was a

challenging difficulty in its universal form. We address a

fixed set of physical commands and a logically organised

natural situation, and go forward an easy, yet productive,

method for sign recognition.

Sakshi Gupta and Sushil Kumar “et .al” [3] Human gesture

recognition was an stimulating research area. Hand gesture

recognition could have marvellous applications in Human

Computer interface .The mouse and keyboard were

presently the main interface between man and computer.

There was a need of mechanized hand that could perform

events like human hand in real time application, as it was not

probable for human to reach up to every object due to not

easy environment. In other areas where 3D series was

required, such as computer games, robotics and design,

other mechanical strategy such as roller-balls, joysticks and

data-gloves were used. User would perform gesture

according to the act as he wanted to be done by robotic hand.

The capability to recognize human gestures open up a broad

range of probable applications such as automatic

identification of sign language to make possible

communication with the hearing impair, using gestures as

input to explain the sentiment of a gesturing person. A

variety of researchers was proposed unlike approach for real

time gesture recognition

3. METHODOLOGY

The methodology is defined as the steps followed for

performing the proposed research work.

METHODOLOGY

Here, the method of the projected work for the hand gesture

recognition system is explained. Firstly the phases of the

hand gesture recognition system are explained and then the

algorithms used in the method are explained. Figure 4.1

explains the methodology, algorithms techniques that are

used to implement this work.

Background Subtraction: The background subtraction

process is the common method of movement of hand

detection. It is a technique that uses the difference of the

current image and the background image to detect the hands

movement region. Its calculation is simple and easy to

implement.




Median filtering: After background subtraction, median

filtering is used to remove noise. Median filter perform 2d

average sifting. The Median Filter block replaces the central

value of an M-by-N neighbourhood with its median value. On

the off chance that the area has a focal point component, the

piece puts the average esteem there. Average separating can

likewise do with the assistance of dialog box. The Main sheet

of the If you choose same as input port I, the output has the

same dimensions as the input to port.

4. EXPERIMENTAL RESULTS

The usefulness and accurateness of any work done could

only be judge by its results and outputs generate. Depending

on the kind of system used and its applications there are

many parameters, basis on which a method is accepted or

rejected. This effectiveness could be calculated only when

the system runs on different datasets and values of different

parameters are recorded and further used to deduce the net

results.

Notwithstanding the way that we are getting ensuring

results with the proposed philosophy, it must be upgraded

for generous data bases. Thus Hand gesture is less

unobtrusive biometric; which offers the possibility to

identify people at a distance, without any interaction or co-

operation from the subject; this is the property which makes

it so attractive.

PARAMETERS USED FOR EVALUATION

As the thesis work is based on matching, the parameters that

could be calculated for evaluating the efficiency of the

system are:

CCR (Correct Classification Rate)

SVM (support vector machine)

Surf Feature

Fig. Various Parameters of input video are calculated

Parameter are calculated in order to extract the four features

of the Hanavan’s model Distance between both hands,

Length of one hand, Length of right hand, Length of left hand,

Height of person .

TRAINED DATABASE RECOGNITION

MATCHING

DATABASE WITH INPUT

TESTING

RESULT

EXPERIMENTAL RESULTS AND CHECK

THE CCRs (CORRECT CLASSIFICATION RATE)

FEATURE EXTRACTION

FEATUREEXTRACTION

BACKGROUND

SUBTRACTION

DATABASE VIDEO AND

CONVERTED INTO FRAMES

INPUTTED VIDEO AND

CONVERTED INTO FRAMES

BACKGROUND SUBTRACTION




Figure: Graph of comparison of pervious work and

proposed work

Figure: Comparison of CCR between previous

and proposed algorithm.

Figure: Graph of comparison of pervious work and proposed

work on the basis of MSE

Figure: Comparison of MSE between previous and proposed

algorithm


work on the basis of PSNR

Figure: Comparison of PSNR between previous and proposed

algorithm


work on the basis of Matching Time.




Figure: Comparison of Matching Time between previous and

proposed algorithm.

5. CONCLUSIONS

With the increasing demands of visual surveillance systems,

human hand identification at a distance has recently gained

more interest. Hand gesture is a potential behavioural

feature and many allied studies have demonstrated that it

has a rich potential as a biometric for recognition. This thesis

has described a simple but effective method for automatic

person recognition from hand silhouette and hand gesture.

Simple feature selection hanavan model reduce the

computational cost significantly during training and

recognition. These methods have been applied on frames of

videos, these videos are live and some from cassia database.

In visual observation frameworks, human ID at a separation

has as of late picked up more investment. The advancement

of workstation vision methods has additionally guaranteed

that vision based programmed motion of hand examination

might be continuously attained. This proposition has

depicted a basic however viable system for programmed

individual recognition from the motion of hand.

REFERENCES [1]D.K. Vishwakarma,Rajiv Kapoor and Rockey Maheshwari

,”An Efficient Approach for the Recognition of Hand

Gestures from Very Low Resolution Images “,2015 Fifth

International Conference on Communication Systems and

Network Technologies, IEEE 2015

[2]Dhanashee Pannasa,”To Analyse Hand Gesture

Recognition for Electronic Device Control”, International

Journal of Advance Research in Computer Science and

Management Studies, Jan 2015

[3] Sakshi Gupta and Sushil Kumar,”Analysis of Hand

Gesture Recognition”, International Journal of

Advanced Research in Computer Science and Software

Engineering, May 2013

[4]V. Padmanabhan and M. Sornalatha,”Hand Gesture

Recognition and Voice Conversion System for Dumb

People”, International Journal of Scientific & Engineering

Research, May 2014

[5]Nikhil Thakur,Sachin Sharma and Gunjan Thakur,

”Review of Hand Recognition Technique”, International

Journal of Advanced Research in Computer Science and

Software Engineering, November 2014

[6]Haitham Hasan and S.Abdul Kareem,” Analysis of Hand

Gesture Recognition by using Human Computer

Interface”, International Journal of Innovation and Scientific

Research, July 2012

[7]Qing Chen,Nicolas Georganas and Emil M Petru, ”Real

Time Vision Based Hand Gesture Recognition Using Haar

Like Features “,Association for Computing Machinery, May

2007.

[8]Joseph .J.Laviola,”A Gesture Recognition as a

Mechanism for Communication with Computer”,

Innovative Space of Scientific Research, June 1999

[9]Deepali N.Kakde,Prof.Dr.J.S Chitode,” Vision Based Real

Time Hand Gesture Recognition System” International

Journal of Advance Research in Computer Science and

Management Studies, Nov 2012

[10]Sidharth S.Routaray and Anupam Aggarwal,”A Review

of Hand Gesture Recognition System”, International

Journal of Innovation and Scientific Research, June 2011.




[11]Pragati Garg, Naveen Aggarwal and Sanjeev Sofat,”3D

Based Approaches of Hand Gesture Recognition System”,

Association for Computing Machinery, June 2013

A Study of Hand Gesture Recognition Technique Engineering College, Mouli (Barwala) ... resolution images has always been the focus in the dispensation of the digital images.

Documents